1 /* internal.h: mm/ internal definitions
3 * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
11 #ifndef __MM_INTERNAL_H
12 #define __MM_INTERNAL_H
16 #include <linux/pagemap.h>
18 extern int do_swap_page(struct mm_struct *mm, struct vm_area_struct *vma,
19 unsigned long address, pte_t *page_table, pmd_t *pmd,
20 unsigned int flags, pte_t orig_pte);
23 * The set of flags that only affect watermark checking and reclaim
24 * behaviour. This is used by the MM to obey the caller constraints
25 * about IO, FS and watermark checking while ignoring placement
26 * hints such as HIGHMEM usage.
28 #define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\
29 __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\
30 __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC)
32 /* The GFP flags allowed during early boot */
33 #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS))
35 /* Control allocation cpuset and node placement constraints */
36 #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
38 /* Do not use these with a slab allocator */
39 #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
41 void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
42 unsigned long floor, unsigned long ceiling);
44 static inline void set_page_count(struct page *page, int v)
46 atomic_set(&page->_count, v);
49 extern int __do_page_cache_readahead(struct address_space *mapping,
50 struct file *filp, pgoff_t offset, unsigned long nr_to_read,
51 unsigned long lookahead_size);
54 * Submit IO for the read-ahead request in file_ra_state.
56 static inline unsigned long ra_submit(struct file_ra_state *ra,
57 struct address_space *mapping, struct file *filp)
59 return __do_page_cache_readahead(mapping, filp,
60 ra->start, ra->size, ra->async_size);
64 * Turn a non-refcounted page (->_count == 0) into refcounted with
67 static inline void set_page_refcounted(struct page *page)
69 VM_BUG_ON_PAGE(PageTail(page), page);
70 VM_BUG_ON_PAGE(atomic_read(&page->_count), page);
71 set_page_count(page, 1);
74 extern unsigned long highest_memmap_pfn;
79 extern int isolate_lru_page(struct page *page);
80 extern void putback_lru_page(struct page *page);
81 extern bool zone_reclaimable(struct zone *zone);
86 extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
93 * Structure for holding the mostly immutable allocation parameters passed
94 * between functions involved in allocations, including the alloc_pages*
95 * family of functions.
97 * nodemask, migratetype and high_zoneidx are initialized only once in
98 * __alloc_pages_nodemask() and then never change.
100 * zonelist, preferred_zone and classzone_idx are set first in
101 * __alloc_pages_nodemask() for the fast path, and might be later changed
102 * in __alloc_pages_slowpath(). All other functions pass the whole strucure
103 * by a const pointer.
105 struct alloc_context {
106 struct zonelist *zonelist;
107 nodemask_t *nodemask;
108 struct zone *preferred_zone;
111 enum zone_type high_zoneidx;
112 bool spread_dirty_pages;
116 * Locate the struct page for both the matching buddy in our
117 * pair (buddy1) and the combined O(n+1) page they form (page).
119 * 1) Any buddy B1 will have an order O twin B2 which satisfies
120 * the following equation:
122 * For example, if the starting buddy (buddy2) is #8 its order
124 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
126 * 2) Any buddy B will have an order O+1 parent P which
127 * satisfies the following equation:
130 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
132 static inline unsigned long
133 __find_buddy_index(unsigned long page_idx, unsigned int order)
135 return page_idx ^ (1 << order);
138 extern int __isolate_free_page(struct page *page, unsigned int order);
139 extern void __free_pages_bootmem(struct page *page, unsigned long pfn,
141 extern void prep_compound_page(struct page *page, unsigned int order);
142 #ifdef CONFIG_MEMORY_FAILURE
143 extern bool is_free_buddy_page(struct page *page);
145 extern int user_min_free_kbytes;
147 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
153 * compact_control is used to track pages being migrated and the free pages
154 * they are being migrated to during memory compaction. The free_pfn starts
155 * at the end of a zone and migrate_pfn begins at the start. Movable pages
156 * are moved to the end of a zone during a compaction run and the run
157 * completes when free_pfn <= migrate_pfn
159 struct compact_control {
160 struct list_head freepages; /* List of free pages to migrate to */
161 struct list_head migratepages; /* List of pages being migrated */
162 unsigned long nr_freepages; /* Number of isolated free pages */
163 unsigned long nr_migratepages; /* Number of pages to migrate */
164 unsigned long free_pfn; /* isolate_freepages search base */
165 unsigned long migrate_pfn; /* isolate_migratepages search base */
166 unsigned long last_migrated_pfn;/* Not yet flushed page being freed */
167 enum migrate_mode mode; /* Async or sync migration mode */
168 bool ignore_skip_hint; /* Scan blocks even if marked skip */
169 int order; /* order a direct compactor needs */
170 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
171 const int alloc_flags; /* alloc flags of a direct compactor */
172 const int classzone_idx; /* zone index of a direct compactor */
174 int contended; /* Signal need_sched() or lock
175 * contention detected during
181 isolate_freepages_range(struct compact_control *cc,
182 unsigned long start_pfn, unsigned long end_pfn);
184 isolate_migratepages_range(struct compact_control *cc,
185 unsigned long low_pfn, unsigned long end_pfn);
186 int find_suitable_fallback(struct free_area *area, unsigned int order,
187 int migratetype, bool only_stealable, bool *can_steal);
192 * This function returns the order of a free page in the buddy system. In
193 * general, page_zone(page)->lock must be held by the caller to prevent the
194 * page from being allocated in parallel and returning garbage as the order.
195 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
196 * page cannot be allocated or merged in parallel. Alternatively, it must
197 * handle invalid values gracefully, and use page_order_unsafe() below.
199 static inline unsigned int page_order(struct page *page)
201 /* PageBuddy() must be checked by the caller */
202 return page_private(page);
206 * Like page_order(), but for callers who cannot afford to hold the zone lock.
207 * PageBuddy() should be checked first by the caller to minimize race window,
208 * and invalid values must be handled gracefully.
210 * READ_ONCE is used so that if the caller assigns the result into a local
211 * variable and e.g. tests it for valid range before using, the compiler cannot
212 * decide to remove the variable and inline the page_private(page) multiple
213 * times, potentially observing different values in the tests and the actual
216 #define page_order_unsafe(page) READ_ONCE(page_private(page))
218 static inline bool is_cow_mapping(vm_flags_t flags)
220 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
224 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
225 struct vm_area_struct *prev, struct rb_node *rb_parent);
228 extern long populate_vma_page_range(struct vm_area_struct *vma,
229 unsigned long start, unsigned long end, int *nonblocking);
230 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
231 unsigned long start, unsigned long end);
232 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
234 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
238 * must be called with vma's mmap_sem held for read or write, and page locked.
240 extern void mlock_vma_page(struct page *page);
241 extern unsigned int munlock_vma_page(struct page *page);
244 * Clear the page's PageMlocked(). This can be useful in a situation where
245 * we want to unconditionally remove a page from the pagecache -- e.g.,
246 * on truncation or freeing.
248 * It is legal to call this function for any page, mlocked or not.
249 * If called for a page that is still mapped by mlocked vmas, all we do
250 * is revert to lazy LRU behaviour -- semantics are not broken.
252 extern void clear_page_mlock(struct page *page);
255 * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
256 * (because that does not go through the full procedure of migration ptes):
257 * to migrate the Mlocked page flag; update statistics.
259 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
261 if (TestClearPageMlocked(page)) {
262 int nr_pages = hpage_nr_pages(page);
264 /* Holding pmd lock, no change in irq context: __mod is safe */
265 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
266 SetPageMlocked(newpage);
267 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
271 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
274 * At what user virtual address is page expected in @vma?
276 static inline unsigned long
277 __vma_address(struct page *page, struct vm_area_struct *vma)
279 pgoff_t pgoff = page_to_pgoff(page);
280 return vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
283 static inline unsigned long
284 vma_address(struct page *page, struct vm_area_struct *vma)
286 unsigned long address = __vma_address(page, vma);
288 /* page should be within @vma mapping range */
289 VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
294 #else /* !CONFIG_MMU */
295 static inline void clear_page_mlock(struct page *page) { }
296 static inline void mlock_vma_page(struct page *page) { }
297 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
299 #endif /* !CONFIG_MMU */
302 * Return the mem_map entry representing the 'offset' subpage within
303 * the maximally aligned gigantic page 'base'. Handle any discontiguity
304 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
306 static inline struct page *mem_map_offset(struct page *base, int offset)
308 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
309 return nth_page(base, offset);
310 return base + offset;
314 * Iterator over all subpages within the maximally aligned gigantic
315 * page 'base'. Handle any discontiguity in the mem_map.
317 static inline struct page *mem_map_next(struct page *iter,
318 struct page *base, int offset)
320 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
321 unsigned long pfn = page_to_pfn(base) + offset;
324 return pfn_to_page(pfn);
330 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
331 * so all functions starting at paging_init should be marked __init
332 * in those cases. SPARSEMEM, however, allows for memory hotplug,
333 * and alloc_bootmem_node is not used.
335 #ifdef CONFIG_SPARSEMEM
336 #define __paginginit __meminit
338 #define __paginginit __init
341 /* Memory initialisation debug and verification */
348 #ifdef CONFIG_DEBUG_MEMORY_INIT
350 extern int mminit_loglevel;
352 #define mminit_dprintk(level, prefix, fmt, arg...) \
354 if (level < mminit_loglevel) { \
355 if (level <= MMINIT_WARNING) \
356 printk(KERN_WARNING "mminit::" prefix " " fmt, ##arg); \
358 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
362 extern void mminit_verify_pageflags_layout(void);
363 extern void mminit_verify_zonelist(void);
366 static inline void mminit_dprintk(enum mminit_level level,
367 const char *prefix, const char *fmt, ...)
371 static inline void mminit_verify_pageflags_layout(void)
375 static inline void mminit_verify_zonelist(void)
378 #endif /* CONFIG_DEBUG_MEMORY_INIT */
380 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
381 #if defined(CONFIG_SPARSEMEM)
382 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
383 unsigned long *end_pfn);
385 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
386 unsigned long *end_pfn)
389 #endif /* CONFIG_SPARSEMEM */
391 #define ZONE_RECLAIM_NOSCAN -2
392 #define ZONE_RECLAIM_FULL -1
393 #define ZONE_RECLAIM_SOME 0
394 #define ZONE_RECLAIM_SUCCESS 1
396 extern int hwpoison_filter(struct page *p);
398 extern u32 hwpoison_filter_dev_major;
399 extern u32 hwpoison_filter_dev_minor;
400 extern u64 hwpoison_filter_flags_mask;
401 extern u64 hwpoison_filter_flags_value;
402 extern u64 hwpoison_filter_memcg;
403 extern u32 hwpoison_filter_enable;
405 extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
406 unsigned long, unsigned long,
407 unsigned long, unsigned long);
409 extern void set_pageblock_order(void);
410 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
411 struct list_head *page_list);
412 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
413 #define ALLOC_WMARK_MIN WMARK_MIN
414 #define ALLOC_WMARK_LOW WMARK_LOW
415 #define ALLOC_WMARK_HIGH WMARK_HIGH
416 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
418 /* Mask to get the watermark bits */
419 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
421 #define ALLOC_HARDER 0x10 /* try to alloc harder */
422 #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
423 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
424 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
425 #define ALLOC_FAIR 0x100 /* fair zone allocation */
428 struct tlbflush_unmap_batch;
430 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
431 void try_to_unmap_flush(void);
432 void try_to_unmap_flush_dirty(void);
434 static inline void try_to_unmap_flush(void)
437 static inline void try_to_unmap_flush_dirty(void)
441 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
442 #endif /* __MM_INTERNAL_H */